REMOTE
SENSINGA G
A
LABORATORY
http://rsl.gis.umn.edu
ND
FACT
SHEET
EOSPATIAL
NALYSIS
UNIVERSITY OF MINNESOTA
~
COLLEGE OF NATURAL RESOURCES
~
DEPARTMENT OF FOREST RESOURCES
Mapping Forest Resources Using the
k-Nearest Neighbor Method
6
Forests play a major economic role in Minnesota, as the forest industry generates annual sales of
$8.5 billion. Minnesota forests also add significant environmental and recreational benefits.
For many years, natural resource professionals have used aerial photography to
inventory and manage forests. Research demonstrates that remote sensing can provide
quantitative information for forest resource inventory and management applications. The
synoptic view of landscapes generated by satellite -borne multispectral sensors provides the
opportunity for applications ranging from stand management to regional planning.
With representatives from government agencies and industry, we are using satellite
remote sensing to map forest resources, improving forest inventory and management at local,
state, and regional scales. Our goal is to develop remote sensing-based approaches to
acquire up-to-date information on Minnesota forests, equipping forestry professionals with the
tools to support management and policy decision making. This fact sheet is an overview of
one such approach called the k -Nearest Neighbor method.
T
FOREST INVENTORY MAPPING
he USDA Forest Service Forest Inventory
and Analysis (FIA) program has been
conducting state-by-state and ultimately
nationwide forest inventories for decades.
However, these field plot-based inventories
have not been able to produce precise county
and local estimates and useful operational
maps.
Additionally, traditional satellite-based forest
classifications have yet to match detailed
forest type identification with ground-based
survey definitions to produce for interpolation
and extrapolation of FIA data. Precise
classification has been limited to general or
aggregate classes of little use to improving
inventory precision and providing truly useful
operational forest maps.
M APPING WITH K-NEAREST NEIGHBOR
Recognizing the great potential of satellite
imagery to help meet the need for wall-to-wall
digital mapping of attributes, the Finnish
Forest Research Institute initiated the Finnish
Multi-Source National Forest Inventory
(FMS-NFI) in 1990. Since this time, FMSNFI has made steady progress towards forest
type and structure estimate mapping. First
introduced to forest inventory practice by
Tomppo (1991), the k-Nearest Neighbor
(kNN) algorithm is today the primary tool
used by the national forest agencies of several
Nordic countries to classify and map a host of
forest attributes.
The kNN method assigns each unknown
(target) pixel of a satellite image the field
attributes of the most similar reference pixel(s)
for which field data exists. The similarity is
defined in terms of the feature space, typically
measured as Euclidean or Mahalanobis
distance between spectral bands. The nonparametric algorithm of this unique method
overcomes low sampling restrictions to
produce truly useful maps of forest attributes
collected through field inventory (FrancoLopez et al., 2001).
Remote sensing and GIS research making a difference to Minnesota’s natural resources
1
FACT S HEET 6
Figure 1. kNN estimate of forest cover type and timber volume in cubic meters with calculations based on 1,835 FIA subplots
and 3 Landsat TM+ dates.
The kNN multi-source inventory has proved
timely, cost-efficient, and accurate in both the
Nordic countries and initial trials in the U.S.
(McRoberts et al., 2002). This approach for
extending field point inventories is ideally
suited to the estimation and monitoring needs of
federal agencies, such as the Forest Service, that
conduct natural and agricultural resource
inventories. It provides wall-to-wall maps of
forest attributes, retains the natural data
variation found in the field inventory (unlike
many parametric algorithms), and provides
precise and localized estimates in common
metrics across large areas and ownerships
(Figure 1).
References
Tomppo, E. 1991. Satellite imagery-based national
inventory of Finland. International Archives of
Photogrammetry and Remote Sensing. 28: 7-1,
419-424.
Franco-Lopez, H., Ek, A.R., and Bauer, M.E. 2001.
Estimation and mapping of forest stand density,
volume, and cover type using the k-nearest
neighbors method. Remote Sensing of
Environment. 77: 3, 251-274.
McRoberts, R.E., Nelson, M.D., and Wendt, D.G.
2002. Stratified estimation of forest area using
satellite imagery, inventory data, and the k-nearest
neighbors technique. Remote Sensing of
Environment. 82: 2-3, 457-468.
For more information, contact:
Alan Ek
University of Minnesota
Department of Forest Resources
1530 Cleveland Ave. N.
St. Paul, MN 55108
Phone: (612) 624-6400
E-mail: [email protected]
Forest mapping research at the University
of Minnesota is supported by:
NASA, the Minnesota DNR , and the
USDA Forest Service
The University of Minnesota’s Remote Sensing and Geospatial Analysis Laboratory (RSL), a unit of the Department
of Forest Resources and College of Natural Resources, was established in 1972 and focuses on geospatial research
and development for forestry and natural resources.
Current efforts emphasize quantitative approaches to natural resource assessment, carried out in cooperation with
resource agencies. Core activities at the RSL include research, education and outreach, and the facilities feature an
array of hardware and software for image processing, mapping, modeling, statistical analysis and visualization.
2
http://rsl.gis.umn.edu